It’s a tool to find meaningful quotes from movies by searching with a movie title and a theme (like “love,” “betrayal,” or “courage”).
The problem it solves: I often wanted to find a specific quote from a movie that captured a certain feeling, but couldn’t remember the exact words. Traditional keyword searches on existing quote databases are too literal. If you search for “love quote from Inception,” you might get nothing, even if the movie has lines deeply related to the concept of love.
How it works (the technical side): Instead of relying on keyword matching, the tool uses an embedding model to convert movie dialogues and the user’s query into high-dimensional vectors. It then performs a semantic similarity search to find the most conceptually relevant lines, even if they don’t contain the exact keyword. The front end is a simple static site built with Next.js, and the search is handled by a serverless function calling a vector database.
Why I built it: This was a weekend project to experiment with applying semantic search to a concrete, everyday problem outside of the typical enterprise or coding contexts (like documentation search). It demonstrates how AI can understand the meaning behind a request rather than just matching words.
I’d be interested in your feedback, especially on:
The accuracy of the quote matches. Ideas for other non-traditional datasets where semantic search could be useful. The overall user experience. Try it out here: https://www.aimoviequotes.com